Abstract:For e-commerce platforms such as Taobao and Amazon, advertisers play an important role in the entire digital ecosystem: their behaviors explicitly influence users' browsing and shopping experience; more importantly, advertiser's expenditure on advertising constitutes a primary source of platform revenue. Therefore, providing better services for advertisers is essential for the long-term prosperity for e-commerce platforms. To achieve this goal, the ad platform needs to have an in-depth understanding of advertisers in terms of both their marketing intents and satisfaction over the advertising performance, based on which further optimization could be carried out to service the advertisers in the correct direction. In this paper, we propose a novel Deep Satisfaction Prediction Network (DSPN), which models advertiser intent and satisfaction simultaneously. It employs a two-stage network structure where advertiser intent vector and satisfaction are jointly learned by considering the features of advertiser's action information and advertising performance indicators. Experiments on an Alibaba advertisement dataset and online evaluations show that our proposed DSPN outperforms state-of-the-art baselines and has stable performance in terms of AUC in the online environment. Further analyses show that DSPN not only predicts advertisers' satisfaction accurately but also learns an explainable advertiser intent, revealing the opportunities to optimize the advertising performance further.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to address the issues of advertiser behavior intention understanding and satisfaction prediction on e-commerce platforms (such as Taobao, Amazon). Specifically: 1. **Advertiser Behavior Intention Understanding**: - Advertisers achieve their marketing goals through various advertising activities, leading to frequent and rich interactions with the advertising platform. The various actions taken by advertisers in advertising campaigns (such as adding or removing tags, adjusting bids, etc.) reflect their marketing intentions. - Traditional research mainly focuses on user behavior modeling and intention recognition, with less attention paid to advertisers' behavior, intentions, or satisfaction. Therefore, this paper proposes a novel approach to identify advertisers' marketing intentions. 2. **Advertiser Satisfaction Prediction**: - Advertiser satisfaction is crucial for the long-term performance of the advertising system. If advertisers are dissatisfied with the advertising results, they may stop advertising, affecting the platform's revenue. Therefore, predicting advertiser satisfaction is important for improving the performance of the advertising system and maintaining advertiser loyalty. - This paper proposes a new metric to evaluate advertiser satisfaction and indirectly measures it through advertisers' investment behaviors (such as changes in advertising spending). ### Main Contributions 1. **First Consideration of Advertiser-Side Intention Recognition and Satisfaction Prediction**: - Defines advertiser intention as a multi-dimensional weight vector, which can be derived through a mathematical optimization problem. - Proposes a new metric to evaluate advertiser satisfaction, based on changes in advertising investment. 2. **Proposes a Deep Satisfaction Prediction Network (DSPN)**: - DSPN is a supervised learning framework with a two-stage hierarchical structure. The first stage learns latent intentions from advertisers' historical data through feature transformation, attention mechanisms, and recurrent neural networks (RNN); the second stage explicitly models the relationship between advertiser intention vectors and satisfaction. - DSPN effectively extracts key information from the interactions between advertisers and the advertising platform, used to understand advertisers' intentions and satisfaction. 3. **Experimental Validation**: - Extensive experiments were conducted on Alibaba's advertising dataset, validating the effectiveness of DSPN in satisfaction prediction and intention vector generation. - The generated intention vectors not only explain advertisers' behaviors well but also reveal potential targets for further optimizing the advertising system's performance. - DSPN has been deployed in Alibaba's online advertiser churn prediction system, serving daily business needs. ### Related Work 1. **User Intention Modeling and Satisfaction Prediction**: - Existing research mainly focuses on predicting user search query satisfaction and user intentions in recommendation systems. - Advertiser intention modeling differs significantly from user intention modeling, as advertiser intentions are more complex and difficult to model directly. 2. **Click-Through Rate Prediction**: - On e-commerce platforms, click-through rate (CTR) is used to measure users' interest and satisfaction with recommended items. - Advertiser satisfaction prediction differs from CTR prediction in three main aspects: label acquisition difficulty, satisfaction complexity, and different objectives. ### Basic Knowledge 1. **Interaction Between Advertisers and Display Advertising Platforms**: - Describes how advertisers launch and close advertising campaigns on the Taobao advertising platform and take various actions in advertising campaigns to achieve their marketing goals. - Introduces the definitions of advertiser intention and satisfaction, and how to establish the connection between them by extracting interaction information. 2. **Advertiser Intention Modeling**: - Evaluates advertiser intentions through surveys and mathematical optimization models. - Introduces an advertiser-specific weight vector to define advertiser intentions, reflecting advertisers' preferences for different metrics. 3. **Advertiser Satisfaction Modeling**: - Indirectly measures advertiser satisfaction through advertisers' investment behaviors (such as changes in advertising spending). - Proposes a satisfaction evaluation method based on investment-related metrics. ### Deep Satisfaction Prediction Network (DSPN) 1. **Input Features**: - Includes advertiser ID features, performance metrics, and action features. - Action features are divided into

A Deep Prediction Network for Understanding Advertiser Intent and Satisfaction

Deep Page-Level Interest Network in Reinforcement Learning for Ads Allocation

Deep Interest Network for Click-Through Rate Prediction

Predict Click-Through Rates with Deep Interest Network Model in E-commerce Advertising

Self-Attention Mechanism Enhanced User Interests Modeling for Personalized Recommendation Services in Cyber-Physical-Social Systems.

Buying or Browsing?: Predicting Real-time Purchasing Intent Using Attention-based Deep Network with Multiple Behavior

ADSNet: Cross-Domain LTV Prediction with an Adaptive Siamese Network in Advertising

Multi-task Based Sales Predictions for Online Promotions

EXTR: Click-Through Rate Prediction with Externalities in E-Commerce Sponsored Search

Know in AdVance: Linear-Complexity Forecasting of Ad Campaign Performance with Evolving User Interest

An adaptive hybrid XdeepFM based deep Interest network model for click-through rate prediction system

Deep Situation-Aware Interaction Network for Click-Through Rate Prediction.

Deep Spatio-Temporal Neural Networks for Click-Through Rate Prediction

EdgeDIPN

We Know What You Want: An Advertising Strategy Recommender System for Online Advertising

LEAD-ID: Language-Enhanced Denoising and Intent Distinguishing Graph Neural Network for Sponsored Search Broad Retrievals

Advertising Click-Through Rate Prediction Model Based on an Attention Mechanism and a Neural Network

Adaptive Domain Interest Network for Multi-domain Recommendation

A two-stage prediction model based on behavior mining in livestream e-commerce

Online Advertising Revenue Forecasting: An Interpretable Deep Learning Approach

Personalized Attraction Enhanced Sponsored Search with Multi-task Learning